New map shows which jobs are most vulnerable to AI

A new project attempts to estimate how vulnerable different professions are to artificial intelligence.

Andrej Karpathy, former AI director at Tesla and one of the founding members of OpenAI, published a project that evaluates the exposure of 342 professions in the US economy to AI.

Each occupation received a score from 0 to 10, where higher numbers indicate greater vulnerability. The results are presented as an interactive treemap: the size of each block reflects the number of workers in that profession, while the color ranges from green (low risk) to red (high risk).

Karpathy later removed the original repository from GitHub for unclear reasons, possibly because the project still requires refinement. However, the code and dataset remain available in a forked version.

Across all professions, the average score is 5.3 out of 10. Software developers received a score of 9, accountants and secretaries 8–9, customer support specialists 9, lawyers 8, while medical transcriptionists were assigned the only perfect score of 10. At the opposite end of the scale are roofers, construction workers, electricians, and plumbers, whose scores range from 0 to 3.

The pattern is straightforward: jobs that produce purely digital output and can be performed remotely on a computer face the highest AI exposure. Roles that require physical presence, manual labor, or direct human interaction currently have a natural layer of protection.